diameter, is much the best form to use, as it gives a large surface for action and a corresponding uniformity in the flow of gas, which can be obtained from no other form. Using the purest zinc obtainable, containing 0.0011 per cent. iron, no arsenic, and leaving no residue or but a mere trace, in the generator, when slight excess of acid was added, we next proceeded to determine the effect of adding known amounts of iron and of ferric chloride. Some 30 to 40 grams of zinc were used for each experiment. In order to obtain an alloy of iron and zine we used the zinc in the rod form, added iron reduced by hydrogen, fused the whole and granulated. The alloying was fairly successful when small amounts of iron were used but was seldom complete with the larger quantities. Each sample made was proved to give no mirror itself before the addition of any arsenic. A preliminary experiment on the purest (Fe 0.0011 per cent.) zinc gave results as follows: Arsenic taken, 0.0050 gram; arsenic found, first mirror, 0.0048 gram. On evaporating with nitric acid and re-treating the small amount of red precipitate formed, a light brown mirror, too small to be weighed, but representing a small fraction of a milligram was obtained. This is especially noteworthy as showing that even with this small amount of iron some arsenic is retained. EXPERIMENTS WITH ZINC ALLOYED WITH IRON. The alloying was not always perfect, especially with the larger amount. 1. Alloy, 100 grams zinc; 0.2 gram iron: Arsenic taken, 0.0025 gram; first mirror, 0.0008 gram, 32 per cent. 2. Alloy, 100 grams zine; I gram iron: Arsenic taken, 0.0050 gram; first mirror, 0.0017 gram, 34 per cent. 3. Alloy, 100 grams zine; 5 grams iron: Arsenic taken, 0.0050 gram; first mirror, 0.0017 gram, 34 per cent. 4. Alloy, 100 grams zinc; 20 grams iron: Arsenic taken, 0.0050 gram: first mirror, 0.0025 gram, 50 per cent. EXPERIMENTS TO SHOW THE EFFECT OF FERRIC CHLORIDE IN THE GENERATOR SOLUTION. 1. 30 grams purest zinc, with 0.2 per cent. iron as FeCl, added to generator: Arsenic taken, 0.0050 gram; first mirror, 0.00345 gram, 69 per cent. 2. 30 grams purest zinc, with 1 per cent. iron as FeCl, added to generator: Arsenic taken, 0.0050 gram; first mirror, 0.0027 gram, 54 per cent. 3. 30 grams purest zinc with 5 per cent. iron as FeCl, added to generator: Arsenic taken, 0.0050 gram; first mirror, 0.00295 gram, 59 per cent. 4. 30 grams purest zinc with 20 per cent. iron as FeCl, added to generator: Arsenic taken, 0.0050 gram; first mirror, 0.0016 gram, 32 per cent. 5. 30 grams purest zinc with 15 per cent. iron as FeCl, aaded to generator: Arsenic taken, 0.0001 gram; first mirror, less than 0.000001 gram. From the results of our work we are forced to the conclusion that iron, whether as an alloy with the zinc or in the generator as a soluble salt, must be avoided, if exact quantitative results are desired. Whenever iron was present, even in small amounts, we have never failed to find arsenic in the red precipitate of hydrated iron oxide, formed by evaporating the liquid contents of the generator with nitric acid. This fact must give rise to serious error in the determination of very small amounts of arsenic by the use of standard mirrors, unless special precautions are taken, and it is for the determination of these small fractions of a milligram that the Marsh-Berzelius method finds its special application. Although our results do not show any definite amounts of arsenic retained by a given percentage of iron present as impurity we are inclined to believe that the amount retained would be fairly definite where the amount of iron present was very small. If this is true, the method of using standard mirrors, so universally adopted, will give results of essential accuracy, if the precaution be taken to always prepare the standards from the same zinc to be afterwards used in the analysis. It is also necessary that the zinc be as nearly free from iron as can possibly be obtained and that soluble salts of iron be kept out of the generator. It, accordingly, will be necessary in preparing solutions containing iron for analysis to reduce. them and distil off the arsenic, as arsenious chloride, before adding to the generator. As a qualitative test we have never failed to get a mirror, even with iron present, when as much as 0.01 mg. of arsenic was added. NEW HAMPSHIRE COLLEGE, DURHAM, N. H., June 1, 1902. OF NICKEL AND COBALT AS THEY EXIST IN AQUEOUS SOLUTION.' BY O. F. TOWER. Received July 2, 1902. IT has been shown in a former paper that the molecular conductivities of aqueous solutions of nickel and cobalt tartrates are exceptionally small, and furthermore that the apparent molecular weights derived from the freezing-point method considerably exceed the molecular weights calculated from the simple formulas of the salts. It was suggested that these unusual results could be accounted for on the assumption of polymerization. The formula, was given as expressing possibly the constitution of a molecule of nickel tartrate. Such a molecule would very likely be much less dissociated in solution than a simple molecule. In order to investigate this problem more fully these same methods have been applied to the tartrates of other metals and to the nickel, cobalt, and magnesium salts of certain other organic acids. These salts were prepared from pure chemicals of standard make. The solutions were made by treating an excess of the carbonate or oxide of the metal with a sufficiently dilute solution of the acid. In a few instances the hydroxide of the metal was employed. On account of the slight solubility of most of these organic salts, the quantity of salt in solution after the acid had become neutralized was almost never equivalent to the quantity of acid taken, because some of the salt was precipitated while the action was going on. It was therefore necessary to determine the amount of salt actually present in the solution in each case. The temperature of the solution has considerable effect on the solubility 1 Read at the Pittsburg meeting of the American Chemical Society. 2 Tower: This Journal, 22, 501 (1900). of these salts. This effect has already been described for nickel and cobalt tartrates, and is similar for the other salts used, although in most cases less pronounced. Heat seemed to decrease the solubility, so that all solutions were made up in the cold or at a temperature not exceeding 50°. The action of the dilute acids. was very slow at these low temperatures. To accelerate it the solution was constantly shaken until the reaction was neutral. This required only about fifteen minutes with magnesia and the freshly precipitated hydroxides of nickel and cobalt, while for the carbonates of nickel and cobalt an hour or more was frequently necessary. Solutions of barium tartrate were prepared by neutralizing a solution of barium hydroxide with tartaric acid. Solutions containing more than I gram-molecule of barium tartrate in 80 liters were supersaturated. Measurements with these supersaturated solutions revealed no abnormal behavior, which is also the experience of others. Attempts were made to prepare solutions of calcium and zinc tartrates. These salts are, however, so insoluble that the solutions obtained were too weak to render the measurements of any value for the purpose of comparison. The measurements of the electrical conductivity of the tartrates reported in my former article were made with a small combination Wheatstone-Kohlrausch bridge only 25 cm. long. All the measurements which follow were made with a meter bridge, which had been carefully calibrated. The conductivities of the tartrates of nickel and cobalt were therefore redetermined with the new appaThe temperature at which all the determinations were made was 18° ± 0.1°. The conductivity of the water used varied from 2.0 to 3.0 X 10-6. This has been deducted from the specific conductivity in every case. The results with the salts of tartaric, malic and succinic acids are given in Table I: v is the number of liters in which a gram-molecule of the salt was dissolved; M is the molecular conductivity in reciprocal ohms. In the former article the equivalent conductivity was given, but since some of the salts probably exist in a polymerized condition, the molecular conductivity is given as affording a better basis of comparison. 1 Tower: loc. cit., pp. 504 and 515. |